The current goal of initial antiretroviral (ARV) therapy is suppression of plasma HIV-1 RNA levels to below the limits of currently available assays. A substantial proportion HIV-1 infected patients who initiate antiretroviral therapy in clinical practice or antiretroviral clinical trials either fail to suppress HIV RNA levels or have HIV RNA levels rebound on therapy. Frequently these patients will have sustained CD4 cell count responses and limited clinical symptoms and therefore have limited clinical indications for altering therapy despite increases in viral replication. On the other hand increased viral replication on therapy leads to selection of resistance mutations to the antiretroviral agents comprising their therapy and potentially cross resistance to other agents in the same class decreasing the likelihood of response to subsequent antiretroviral therapy. The optimal timing of switching antiretroviral therapy to ensure virologic suppression and prolonged clinical stability in patients who have rebound in their viral load, yet are stable clinically, is not known. Randomized clinical trials to compare early versus delayed switching have been difficult to design and more difficult to enroll. Other means of exploring this question are needed. In recent ACTG clinical trials, such as ACTG A5095, patients randomized to initial antiretroviral treatment combinations, who fail to suppress HIV-1 RNA or have a rebound of HIV RNA on therapy are allowed, via study protocol, to switch the ARV to which they were randomized in favor of another ARV. The primary goal of the new study design is to achieve viral load suppression. However the timing of the switch to a new regimen is not dictated by the protocol. The natural question then arises, when should patients or their attending physicians switch ARVs in order to reduce the average time to viral load suppression and to prolong the duration of viral load suppression on the subsequent regimen? This new study design requires a detailed record of data leading up to that time when patients switch treatment which allows investigators to address the question posed above in ways that were unavailable in earlier studies. We propose new methods which model the clinical endpoint, e.g. time to viral load suppression, as a function of switching time. We also consider scenarios where switching time may be censored due to switch-precluding events, in which case a patient intended to switch but the actual switching time and second ARV are unobserved. We term the sequence of decision rules, switch to a new ARV a at viral load / given the past history of treatments and covariate information, a sequential antiretroviral treatment regimen. [unreadable] [unreadable] [unreadable]